Introduction

Preterm delivery (PTD), defined as delivery prior to thirty-seven weeks gestational age, is a marker of obstetrical inequity and the leading cause of infant mortality [1–4]. Between 2011 and 2015, about eight percent of rural labor and delivery units closed their doors due to poor reimbursement policies, high costs of childbirth, low patient volume, and the lack of rural providers [5]. As of 2014, more than half of rural counties in the Unites States were described as maternity care deserts, with limited prenatal care and no hospital-based labor and delivery services [5]. Rural counties with a larger population of non-Hispanic Black women of reproductive age had higher odds of closure, exacerbating existing disparities in maternal and infant outcomes for rural Black women when compared to their rural White counterparts [5]. In the first year after the loss of hospital–based obstetric services in rural counties, there was an increased likelihood of having both nonlocal and preterm births, increased maternal mortality, and increased rural infant mortality rates [6–8]. In New Hampshire, nine out of sixteen rural hospitals have closed since the year 2000, with two units closing in early 2018 [9]. Such closures have increased travel times, worsened all-cause birth outcomes, and decreased frequency of prenatal visits [9], especially for patients using Medicaid (S1 Fig). Medicaid covers nearly half of all US rural births; poor reimbursement packages to rural hospitals have furthered the progression of unit closures [5].

Interestingly, a deluge of rural hospital closures in the 1980s led to the development of critical access hospital designations by Centers for Medicaid and Medicaid Services (CMS); this designation was made to reduce the financial vulnerability and improve rural hospital access. However, there is no requirement for Critical Access Hospitals to include labor and delivery units. This disparity in policy allows for continued closure of these maternity care units. Recent recommendations from the Alliance on Innovation on Maternal Health (AIM) include conducting internal institutional data review to assess obstetrical outcomes and identify gaps in rural maternal health equity. While there are several studies assessing the impact of closures on preterm birth incidence, there are no current studies assessing the impact of these successive closures on birth outcomes at a single rural academic tertiary care center. The objective of this study was to examine associations between clinical, sociodemographic, and geographic risks factors with preterm birth, low birth weight, and neonatal intensive care unit admissions at a single rural level III Maternal Obstetric Care unit serving northern New England using multivariable modelling. We hypothesize that insurance carrier status—a proxy for socioeconomic status—and increasing distance from the hospital will be associated with an increased odds of all three outcomes in a rural population. Given the changing obstetrical landscape of rural New Hampshire and current AIM recommendations, analyzing preterm birth delivery offers an essential analysis to identify patients at risk for poor birth outcomes, to identify local barriers to access, and to inform institution-wide initiatives to address rural maternal equity at our institution.

Materials and methods

This study was a retrospective electronic medical record cohort study of singleton pregnancies delivering at a single rural academic institution in the US between 2016 and 2018 for analysis between 2020 and 2021. This two-year period coincides with the implementation of ICD-10 diagnostic coding and occurs after the closure of seven regional obstetrical units in New Hampshire [9–11]. No OB units were closed during the study period, offering a period of stability to assess associations. Inclusion criteria includedthose with spontaneously conceived, singleton pregnancies at or greater than 20 weeks gestation delivering between 1/1/2016 to 12/31/2018. Exclusion criteria included multiple gestation, delivery prior to 20 weeks gestational age, and use of in-vitro fertilization for pregnancy (given that it is a confounder for preterm birth). Records were identified for review through a list generated by the Analytics Institute based on claims data (ICD-10 and current procedural terminology (CPT) codes) at a single institution. Chart review was performed by two reviewers to collect all pertinent information and verify billing data. Demographic, clinical (ICD-10 codes), and geographic data were abstracted from these charts (S1 Table).

After extensive literature review, these covariates were determined as impactful on preterm delivery.

Results

Discussion

Among births at our academic medical center, after controlling for transfer status, patients living 1-to-1.5 hours from the hospital had an increased risk for low birth weight, neonatal intensive care unit admission, and preterm birth, compared to those traveling under 30 min (S4 Table). A history of preterm birth and limited prenatal visits were also associated with an increased likelihood of all three adverse outcomes, consistent with previously published data on increased rurality and associated increased preterm birth rates (S4 Table). Although the degree of social deprivation increased in this population with increased distance and increased degree of rurality (S3 and S4 Figs), the degree of social deprivation and insurance carrier status—proxies for socioeconomic deprivation—were not associated with preterm birth, low birth weight or ICN admission in our population. Interestingly, it is the degree of rurality—the RUCA code—that attenuates the relationship between distance to the hospital and preterm birth rates, with statistically significant increases in odds ratios notes for all three outcomes when RUCA codes alone were applied (S6 Table). Rurality is a strong predictor on the risk of all cause preterm birth at our institution, emphasizing the importance of further study into the diverse aspects of New Hampshire rurality impacting this outcome.

Several studies have assessed the relationship between distance and rural maternal health outcomes. Grzybowski et al noted that the adjusted odds ratio for perinatal mortality for newborns from rural communities living greater than four hours away from the hospital was three times greater than those living closer to the hospital [16]. Other studies assessed area level deprivation and its association to maternal morbidity; While the studies looking at effect of distance on rural maternal outcomes determined a significant impact on rates of preterm birth [4,6], neither study explicitly measured the distance from the hospital and its direct correlation to PTB. When examining the distribution of travel times by rurality, most of the area within thirty minutes of the hospital of interest within this study was described as “large rural” per RUCA codes. Yet, patients living an hour away lived in urban, isolated, small rural and large rural communities and were at increased risk for all outcomes. This suggests that distance from the hospital and its impact on preterm deliveries, low birth weight, and ICN admissions is not based solely on degree of rurality, but also on how rurality impacts maternal care, such as road access to the hospital, to local resources, and to community and clinical support. Several of the hospital closures were within one to two hours distance from our academic institution, raising concern that proximity to hospitals is a major risk factor for poor obstetrical outcomes—a concern that is increasing as more hospital units close their obstetrical units [9]). Zip code tabulations are used for geographical factors, but do not capture cohesive areas and often mask the variety of resources, needs, and cultures living in these areas [17] Census tracts have also been used to measure rurality, but it does not follow city or county boundaries and thus runs the risk of rural-urban misclassification [17]. Area level deprivation indices do not always include covariates relevant to rural maternal care [17]. Defining rurality remains an epidemiological challenge and deserves further research to determine the specific resources and characteristics impacting rural maternal care [17].

Strengths of this study include creating local data to compare national and state-wide trends and including multiple variables to assess socioeconomic vulnerability, such as SDI, insurance, or distance. Our data is consistent with findings in published research which lends a level of generalizability to state wide data [18]. Limitations include the potential for unmeasured confounding due to missing covariates in the models and low power for each of the outcomes when divided by distance (S5 Table). Retrospective data collection is limited by inaccurate diagnostic coding and limited numbers of patients from remote areas of the state, impacting the selection and power of the covariates used in the models. For example, inclusion of other substance use disorders beyond opioid use would have enhanced clinical relevancy. Determining which preterm births were spontaneous versus medically indicated—a clinically important distinction—was impossible due to the generalized coding schema used and the retrospective nature of this study. Future studies should consider repeating this methodology at several rural institutions over time to improve the power of covariates and to consider prospective trials in the identified areas of high all-cause preterm birth rates to determine local causes for spontaneous preterm birth specifically.

One in three birthing people now live in an obstetric desert, and more than 50% of rural counties have no hospital-based obstetrical services, forcing rural birthing people to travel more than thirty minutes for care [19]. Hostetter and Klein describe grants, such as the Rural Maternity and Obstetrics Management Strategies, to create regional networks of care for rural people, with networks focusing on the specific cultural, socioeconomic, and geographic needs of their populations [19]. Understanding these granular differences and needs requires institutional data analysis; the Alliance for Innovation on Maternal Health recommends this institutional review of obstetrical data but offers little guidance on recommended best practices to collect and analyze this information. Despite having better socioeconomic deprivation scores, patients traveling from one to one and a half hours away seemed to have increased risk for preterm birth, low birth weight, and neonatal care unit admissions; this indicates the importance of conducting local needs assessments, developing rural maternity health indices to capture identified factors missed in the SDI measurement, and repeating these studies over time to determine if local interventions are improving outcomes for rural maternal health equity and safety. Our study offers a methodologic framework for the collection and analysis of birth outcomes data for a single institution, with the hopes of encouraging specific solutions for the specific local needs of highlighted communities. By identifying the patient catchment areas with high degrees of preterm birth, low birth weight and NICU admission, we hope to be able to be able to provide information for future clinical interventions in the setting of the changing rural maternal health landscape and present a way for individual institutions to assess their own outcomes for future quality improvement.

Supporting information

Acknowledgments

The authors would like to acknowledge Daniel Gelb for his assistance with calculating travel times and Dr. Rebecca Emeny for her editing support.